BVA-S&D-2026-A02

Intelligent Document Processing for Beurtvaartadres | EasyData B.V.
BVA-S&D-2026-A02 · Digitalization: From Paper to Data

From paper to trusted, structured data

EasyData delivers a scalable, API-driven IDP solution that integrates seamlessly into Beurtvaartadres’ target architecture. Secure, transparent, and fully under your control.

Contact the EasyData team →
Supply chain logistics and document processing
25+years in document automation
100+organizations served

ESV Platform: the IDP engine for your data ecosystem

EasyData proposes the ESV Platform as the core IDP engine, integrated into BVA’s target architecture as a pluggable, API-connected module within the Systems & Applications layer. Our solution strictly adheres to BVA’s architectural principles: open APIs, pluggable services, and full data portability. The platform is delivered under a REST-based, API-first design philosophy — all integration uses open formats (JSON, XML) with no proprietary lock-in, ensuring BVA retains full governance over API specifications and integration documentation.

The platform is built on a distributed, multi-tier architecture designed for reliability and horizontal scalability. Documents enter through multiple ingestion channels and are routed via a central Application Server to a Processing Server with built-in load balancing. This processing layer distributes work across multiple parallel Processing Stations, allowing throughput to scale linearly by simply adding stations. The architecture supports growth from 1 million to over 30 million documents per year without re-architecture or forced re-platforming, covering BVA’s full trajectory from Startup through Scale to Heavy Load — including peak handling of 250,000 documents per day with 2x burst capacity to 500,000 documents per day and 2,000 concurrent users.

A dedicated Database Server (MS SQL / PostgreSQL, TCP 1433/5432) manages all metadata, extraction results, and audit trails. Full audit trail retention is maintained across all processing stages, providing complete traceability from document ingestion through structured data output. All processing occurs within Dutch/EU datacentres, ensuring 100% European data sovereignty — data never leaves EU borders. Encryption is applied end-to-end using TLS 1.3 for data in transit and AES-256 for data at rest, meeting the highest security standards. EasyData operates according to ISO 27001 (formal audit March 2026) and is NIS2 compliant, with its own secure EU mail server eliminating dependency on American cloud providers subject to the CLOUD Act.

The web-based EasyVerify Verification System provides browser-based human-in-the-loop review without client installation. Documents falling below configurable confidence thresholds are automatically routed to EasyVerify for manual review according to the four-eyes principle. Operator feedback from EasyVerify flows directly back into the ML model retraining cycle — quarterly during Startup, monthly during Scale, and on-demand at Heavy Load — creating a continuous accuracy improvement loop. Real-time monitoring and operational dashboards powered by Grafana provide full transparency into accuracy trends, exception rates, and system health, with automated alerting when any KPI metric drops more than 2 percentage points below agreed targets over a rolling 30-day window.

One platform, multiple AI engines — always the right technology for the job. The ESV Platform intelligently combines world-leading OCR, open-source computer vision, and premium AI to deliver maximum accuracy at the lowest cost per document.

Business meeting at logistics docks
⬆ AUTO-SCALING ESV PLATFORM · 1M → 30M+ DOCUMENTS/YEAR · THROUGHPUT-DRIVEN · UPWARD & DOWNWARD ⬇ DOCUMENT INPUT ESV PLATFORM CORE · APP SERVER HTTP 80 DB TCP 1433|5432 · LICENSING TCP 3000 FOUR-EYE PRINCIPLE · HUMAN IN THE LOOP ML TRAINING MONITORING · KPI DASHBOARDS · PERFORMANCE DRIFT User Computer BVA Staff / Operators Email / SMTP Secure EU Mail 📁 File Upload Shared / Watch WEB Verif. System 📊 WEB Admin Reporting Console REST API · JSON / XML · HTTP 80 API-first · no vendor lock-in 🖥 Application Server HTTP 80 · IDP Orchestration ISO 27001 · NIS2 🗄 DB Server TCP 1433 (MSSQL) TCP 5432 (PostgreSQL) 🔑 Licensing Server TCP 3000 ABBYY FineReader IDP PROCESSING PIPELINE 📥 1. Ingest Classify 🔍 2. OCR ABBYY Engine 🧠 3. AI/ML Extract + Classify 4. Validate Business Rules 📤 5. Output JSON/XML → BVA STP (Straight Through Processing) · Confidence-based routing Low confidence → EasyVerify Processing Server + Load Balancer Multi-core parallel · 64-bit Routes to Processing Stations 🗂 Processing File Storage EU datacenter only GDPR compliant SCALABLE PROCESSING STATIONS (Throughput-Driven Auto-Scaling) Stage 1 Startup 🖥 1M docs/yr 1–2 stations Jun–Aug 2026 Stage 2 Growth 🖥🖥 5M docs/yr 2–4 stations Nov 2026–Q1 2027 Stage 3 Enterprise 🖥🖥🖥 30M+ docs/yr 4–8+ stations Q2 2027+ ⬇ downscale No re-architecture required at any scale level. Cost matches usage. SLA · RTO/RPO · DR | Business Continuity | Incident Management ISO 27001 (audit Mar 2026) · NIS2 Compliant · 100% Dutch Datacenter · GDPR No data leaves EU borders · Own secure EU mail server · CLOUD Act independent EasyVerify Web Application Browser-based · No client install 👤 Operator 1 Eye #1 Initial Verification Review · Correct Approve / Reject 👤 Operator 2 Eye #2 Independent Verification Counter-check Approve / Reject 👁 👁 Four-Eye Principle Both operators must approve → STP Conflict / Escalation Supervisor review → decision log Verified Result Approved data → Output pipeline JSON/XML → BVA Data Platform ML Feedback Loop Corrections routed to ML Training Accuracy improvement · Model refinement EasyVerify KPIs Avg verification time per doc Correction rate · Override rate Operator throughput metrics GDPR · Audit trail · Access control Role-based auth · Session logging 🤖 ML Training Module Custom model training Computer Vision + LLM 💾 Training Dataset EasyVerify corrections Labelled ground truth EU-only storage 🔄 Model Update Re-train · Test · Deploy Accuracy drift correction Continuous improvement Updated models → OCR/AI engine 📈 Grafana Real-time dashboards Online KPI display Operational transparency 24/7 live monitoring KPI Targets OCR Accuracy ≥ 99% Field Extraction ≥ 95% STP Rate ≥ 80% Proc. time < 60s/doc Uptime SLA ≥ 99.5% DR RTO ≤ 4h 🎯 Drift Detection Accuracy trend analysis Drift indicators + alerts Per doc-type, per stage Triggers ML retraining 📋 Reports Quarterly tactical review Semi-annual strategic eval Exception summaries Segmented by doc type Admin & Reporting Console · WEB-based · Role-based access SLA reporting · Audit trail · Compliance logs · Export to BVA data platform Drift alerts trigger automatic ML retraining pipeline Document Flow IDP Processing Human Verification (Four-Eye) ML Feedback Loop Monitoring / KPI Input Zone ESV Core (Auto-Scaling) EasyVerify HITL ML Training Module Monitoring / KPI Dashboards EasyData B.V. · BVA-S&D-2026-A02 · March 2026
ESV Platform — Enhanced Architecture Diagram · EasyData B.V. · BVA-S&D-2026-A02 · March 2026
Document Input
ESV Core Processing (Auto-Scaling)
Four-Eye Human-in-the-Loop
ML Training Module
Monitoring & KPI Dashboards

11 stages from document to structured data

The ESV Platform processes every document through a transparent, fully auditable pipeline. Each stage is independently configured, monitored, and scalable.

1

Document Ingestion

Multi-channel intake: scans, PDFs, emails, batch uploads, and API submissions. Virus/malware scanning, file integrity checks, and format validation included.

Data Capture →
2

Pre-processing & Normalization

De-skewing, noise removal, image enhancement, and resolution normalization using ABBYY FineReader Engine, OpenCV, YOLO and proprietary algorithms.

Image Enhancement →
3

Document Classification

AI/ML and template-based document type recognition and routing. Supports CMRs, transport documents, declarations, and certificates with confidence scoring.

Document Classification →
4

Data Extraction

Structured extraction via multiple OCR engines with voting mechanism (ABBYY, PaddleOCR, EasyOCR, Azure AI). ICR/HTR for handwriting, ML field extraction, and table detection.

OCR Software →
5

Confidence Scoring

Field-level and document-level reliability assessment with threshold-based routing: automatic pass-through for high-confidence results, referral to human review otherwise.

Data Validation →
6

Validation & Business Rules

Mandatory field checks, format validation (dates, IDs), cross-field consistency checks, and reference data verification via configurable business rule engine.

Data Validation →
7

Human-in-the-Loop

EasyVerify provides web-based review queues with the four-eye principle, manual correction interfaces, and full audit logging — no client installation required.

Human-in-the-Loop →
8

Data Enrichment

Optional enrichment via code normalization (countries, Incoterms, port codes), lookup-based enrichment, and reference matching through configurable pipelines.

Data Processing →
9

Output & Packaging

Structured output in JSON or XML, including document metadata, confidence scores, and validation flags. Canonical data models for logistics domains.

XML Conversion →
10

Audit, Logging & Traceability

Full processing logs, versioned extraction models, who/when/what audit trails, and retention aligned with contractual requirements. Real-time dashboards via Grafana.

Grafana Monitoring →
11

Secure Delivery

API-based delivery to the BVA platform (REST API preferred), secure file transfer (SFTP, HTTPS), and optional event publication via webhooks.

ESV Platform →

What sets the ESV Platform apart

🔌

Open API Architecture

All interaction via REST APIs with JSON/XML output. No proprietary interfaces, no vendor lock-in.

ESV Platform →
👁

Four-Eye Principle

EasyVerify: web-based verification with dual control, escalation management, full audit logging.

EasyVerify →
🤖

Continuous ML Improvement

EasyVerify corrections feed the ML Training Module. 25+ year ABBYY partnership for world-leading OCR/AI.

Machine Learning →
📊

Real-time KPI Dashboards

Grafana dashboards for 24/7 monitoring. Automated alerting when KPIs drop 2pp below target.

Grafana →
🔒

Data Sovereignty & GDPR

100% Dutch/EU datacenter (Kolo, Apeldoorn). No CLOUD Act. Own secure EU mail server.

Data Security →
🔍

Multi-Engine OCR Voting

ABBYY, PaddleOCR, EasyOCR, Azure AI. Multiple engines vote for highest accuracy. 190+ languages.

OCR Comparison →
📄

Logistics Documents

CMRs, customs declarations, certificates, invoices. Canonical models for Incoterms, port codes.

Transport Documents →
🔄

Full Data Portability

BVA retains full ownership. Export in open formats anytime. Cooperative transition, no exit penalties.

Digital Independence →

Transparent Pricing

Pay-per-document, decreasing rate. No hidden costs. Each document charged once. Full downward scalability.

Transparent IT Costs →

Ready to go from paper to data?

Experience our claims firsthand! Request a complimentary Proof of Concept to validate accuracy, throughput, and integration fit within your environment

Contact the EasyData team →